Engineers who run async workflows - what should we build?

We recently pivoted into Python framework for writing arbitrarily complex cloud compute that we called multinode.dev (see the website for detailed docs). The problem we started with was actually how hard it was to build user-triggered asynchronous distributed workflows with seamless autoscaling when working on our previous product, basically something equivalent to https://multinode.dev/docs/core-concepts/job#distributing-a-job-workload-using-functions But it ended up being general end-to-end compute framework and now we're not sure if it was the best idea. We're considering removing service, daemon, and scheduled task in favour of just focusing 100% on job and function primitives that then are deployed with automatically provisioned API (so https://multinode.dev/docs/core-concepts/job#implementing-an-api-for-managing-job-lifecycles is abstracted away) I know it's extremely specific but we would appreciate any feedback we can get from engineers on the site!